Re: Electrolysis update: out-of-process plugins and more

Robert O'Callahan wrote:
> On 17/6/09 10:49 AM, Jeff Muizelaar wrote:
>> On 16-Jun-09, at 2:55 PM, Robert O'Callahan <[email protected]> wrote:
>>> On 16/6/09 6:23 AM, Jeff Muizelaar wrote:
>>>> Shouldn't windowless plugins be easier to support then windowed 
>>>> ones? Is
>>>> there any reason you can't just create a surface in a region of memory
>>>> shared between the plugin process and the Gecko process?
>>>
>>> Yes, there is a reason. Windowless plugins require an X drawable to be
>>> passed into them.
>>
>> Oh. Well in that case things should be even easier. Just ship a drawable
>> id to the plugin process and have it draw to that just like you would a
>> window id.
> 
> Do you mean "whenever we paint our content window, ship the drawable ID 
> to the plugin process and have it synchronously draw to that"? That 
> would work, but it would lose some of the benefits of out-of-process 
> plugins; a slow or hung plugin would retard drawing of the window. The 
> overhead of the synchronous IPC would also impact painting.
> 
> Or do you mean "have the plugin process paint to an X pixmap 
> asynchronously, and copy that pixmap whenever we paint our content 
> window"? That would work, but for transparent plugins you actually need 
> two pixmaps, one with a white background and one with a black 
> background, so you can compute the plugin's alpha values.
> 
> We discussed this at length on IRC a while back. I guess no-one took 
> notes, oops.
> 

I saved the raw IRC conversation, but didn't extract clean notes from 
it.  Here's the raw conversation:

   http://pastebin.mozilla.org/657908

These notes will probably materialize near 
http://wiki.mozilla.org/Content_Processes, but they're not my top 
priority at the moment.

Cheers,
Chris